This definitely seems like an improvement over previous versions. It can now (at least in some cases) generate correct text for a given image. For example the p

This definitely seems like an improvement over previous versions. It can now (at least in some cases) generate correct text for a given image. For example the prompt 'Neon sign saying "Scotland"' generated this https://www.bing.com/images/create/neon-sign-saying-22scotland22/6517fc53711f4fcab4e7d5748498721b?id=RpfDQ3BKyzTqhxnzNiffZA%3d%3d&view=detailv2&idpp=genimg&FORM=GCRIDP&mode=overlay https://www.bing.com/images/create/neon-sign-saying-22scotla... it's still far from perfect though (it struggled with less common words like Kubernetes) but a step in the right direction.